TosicLiuZhu-CASCON2007-FinalSlides.pptts.data61.csiro.au/publications/nicta_full_text/696.pdf ·...
Transcript of TosicLiuZhu-CASCON2007-FinalSlides.pptts.data61.csiro.au/publications/nicta_full_text/696.pdf ·...
7/21/2009
1
CASCON 2007 Presentation by Tosic, Liu, and Zhu The imagination driving Australia’s ICT future
Slide 1 of 24
Towards Business Driven Autonomic
Service-Oriented Computing
Vladimir Tosic (NICTA, UWO, and UNSW)
Yan/Jenny Liu (NICTA and U. of Sydney)
Liming Zhu (NICTA and UNSW)
CASCON 2007 Presentation by Tosic, Liu, and Zhu The imagination driving Australia’s ICT future
Slide 2 of 24
About NICTA
• AustralianAustralianAustralianAustralian center of excellence in useuseuseuse----inspired ICTinspired ICTinspired ICTinspired ICT(info & communications technology) researchresearchresearchresearch
•Established 2002, now 5 labs in 4 cities
•NICTA members (governments and universities):
• NICTA partners:
• 422 research staff, 296 postgraduate students
• 4 research themes (e.g., managing complexity) & 6 business areas (e.g., software infrastructure)
CASCON 2007 Presentation by Tosic, Liu, and Zhu The imagination driving Australia’s ICT future
Slide 3 of 24
Introduction and Motivation
•WSWSWSWS----Policy4MASCPolicy4MASCPolicy4MASCPolicy4MASC is a policy-based language for run-time XML Web service management– PowerfulPowerfulPowerfulPowerful support for specification of business values business values business values business values and strategiesand strategiesand strategiesand strategies for maximization of business values
– In practice (not theory), limited to Web serviceslimited to Web serviceslimited to Web serviceslimited to Web services
• Adaptive Server Framework (ASF)Adaptive Server Framework (ASF)Adaptive Server Framework (ASF)Adaptive Server Framework (ASF) is an autonomic middleware platform for adaptive applications– Policy specificationPolicy specificationPolicy specificationPolicy specification is relatively simple/limitedlimitedlimitedlimited
• AddingAddingAddingAddingWS-Policy4MASC support to ASF– Broadens applicabilityBroadens applicabilityBroadens applicabilityBroadens applicability of WS-Policy4MASC and improves autonomic computing supportimproves autonomic computing supportimproves autonomic computing supportimproves autonomic computing support
– Provides ASF powerful policy specification and support powerful policy specification and support powerful policy specification and support powerful policy specification and support for businessfor businessfor businessfor business----driven managementdriven managementdriven managementdriven management
CASCON 2007 Presentation by Tosic, Liu, and Zhu The imagination driving Australia’s ICT future
Slide 4 of 24
Outline
• Introduction and motivation
• =>=>=>=>WS-Policy4MASC
• Adaptive Server Framework (ASF)
• Adding WS-Policy4MASC support to ASF
• Conclusions and future work
CASCON 2007 Presentation by Tosic, Liu, and Zhu The imagination driving Australia’s ICT future
Slide 5 of 24
WS-Policy4MASC – Background
•Execution of IT systems should be alignedalignedalignedaligned with businessbusinessbusinessbusiness objectives, values, and operations– IT improvements need not increase business value
•Our focus: Web servicesWeb servicesWeb servicesWeb services and their compositions– Distributed components using XML, SOAP, WSDL
•Competitive advantage: specifying and processing business value and business strategybusiness value and business strategybusiness value and business strategybusiness value and business strategy information– Run-time specification: WSWSWSWS----Policy4MASCPolicy4MASCPolicy4MASCPolicy4MASC
– Run-time management: MASC middleware
– Design-time: UML profiles for WS-Policy4MASC
– Run-time selection: UDDI extensions (ongoing)
CASCON 2007 Presentation by Tosic, Liu, and Zhu The imagination driving Australia’s ICT future
Slide 6 of 24
WS-Policy4MASC
• PolicyPolicyPolicyPolicy – high-level operation & management goals
and/or rules (e. g., for security, performance, …)
•WSWSWSWS----PolicyPolicyPolicyPolicy is a general, W3C standardized policy
specification framework for Web services
– Leaves details for extensions (e.g., WS-Security)
•WS-Policy4MASC adds new policy assertions and new policy assertions and new policy assertions and new policy assertions and
detailsdetailsdetailsdetails necessary for run-time management– Feasibility demonstrated by using it in the MASC MASC MASC MASC middlewaremiddlewaremiddlewaremiddleware (see Erradi et al. at ICWS 2007)
– Expressiveness, effectiveness, and usefulness evaluated on 2 case studies (1 simple, 1 medium)
7/21/2009
2
CASCON 2007 Presentation by Tosic, Liu, and Zhu The imagination driving Australia’s ICT future
Slide 7 of 24
WS-Policy4MASC Main Concepts
CASCON 2007 Presentation by Tosic, Liu, and Zhu The imagination driving Australia’s ICT future
Slide 8 of 24
Utility Policy Assertions
• Diverse business valuesDiverse business valuesDiverse business valuesDiverse business values assigned to conditions
– E.g., if (*) was met, requester pays provider $2; otherwise provider pays requester $1
– WS-Policy4MASC advantage is in modeling not only not only not only not only financialfinancialfinancialfinancial business values (prices, penalties), but also but also but also but also many othersmany othersmany othersmany others (e.g., customer satisfaction)
• Business values classified along several dimensionsclassified along several dimensionsclassified along several dimensionsclassified along several dimensions
– Benefits vs. costs, agreed vs. possible, tangible vs. intangible, absolute vs. relative
• Combinations determine business value typesbusiness value typesbusiness value typesbusiness value types
– E.g., absolute intangible possible benefits (models the major aspects of customer satisfaction)
CASCON 2007 Presentation by Tosic, Liu, and Zhu The imagination driving Australia’s ICT future
Slide 9 of 24
Meta-Policy Assertions
•Specify which policy assertions are conflictingpolicy assertions are conflictingpolicy assertions are conflictingpolicy assertions are conflicting and which policy conflict resolution strategypolicy conflict resolution strategypolicy conflict resolution strategypolicy conflict resolution strategy to use
– Policy conflictPolicy conflictPolicy conflictPolicy conflict: 2 or more policy assertions are triggered, but only 1 can be executed (e.g., skip activity vs. replace activity)
•WS-Policy4MASC advantage is in modeling policy conflict resolution strategies that maximize sums strategies that maximize sums strategies that maximize sums strategies that maximize sums of various business value typesof various business value typesof various business value typesof various business value types associated with consequences of choosing 1 of the policies– Model diverse business strategies (e.g., exceptional customer satisfaction)
– MASC implements corresponding algorithms
CASCON 2007 Presentation by Tosic, Liu, and Zhu The imagination driving Australia’s ICT future
Slide 10 of 24
Policy Conflict Resolution
•Strategies are also classified along several classified along several classified along several classified along several dimensionsdimensionsdimensionsdimensions, depending on business value classes– E.g., ‘tangible-only’ vs. ‘intangible-only’ vs. ‘tangible+intangible’
•When difference between summary business values for conflicting policies is smaller than some threshold, tiebreaking rulestiebreaking rulestiebreaking rulestiebreaking rules can be specified– E.g., ‘tangible+intangible’ instead of ‘tangible-only’
•Combinations determine strategiesstrategiesstrategiesstrategies– E.g., ‘intangible-only agreed+possible benefits+ costs with tiebreaking tangible+intangible’ (models maximization of customer satisfaction)
CASCON 2007 Presentation by Tosic, Liu, and Zhu The imagination driving Australia’s ICT future
Slide 11 of 24
MASC Middleware Architecture
CASCON 2007 Presentation by Tosic, Liu, and Zhu The imagination driving Australia’s ICT future
Slide 12 of 24
Some WS-Policy4MASC Limitations
•Supported only by MASC middleware, so limited in limited in limited in limited in practicepracticepracticepractice by MASC characteristics:
– Current applications only for Web servicesWeb servicesWeb servicesWeb services
– MASC prototype is .NET.NET.NET.NET----centeredcenteredcenteredcentered
– MASC has autonomic computingautonomic computingautonomic computingautonomic computing features, but many improvements in this direction are possible
• On the other handOn the other handOn the other handOn the other hand, WS-Policy4MASC (in principle):
– Can be easily applied to other domains
– Is independent of the implementation platform
– Is intended for business-driven management with minimal human intervention (autonomic)
7/21/2009
3
CASCON 2007 Presentation by Tosic, Liu, and Zhu The imagination driving Australia’s ICT future
Slide 13 of 24
Adaptive Server Framework (ASF)
•Developing adaptive systems is difficultdifficultdifficultdifficult– Adaptive logic is determined by application’s requirements for adaptation (applicationapplicationapplicationapplication----specificspecificspecificspecific)
– Developers need appropriate support
• ASF provides basic components and services for developing customized adaptive applicationscustomized adaptive applicationscustomized adaptive applicationscustomized adaptive applications– Application-specific and platform-specific adaptability
– Separation between implementation of adaptive behavior from server application business logic
– Layered architecture (supports legacy systems)
– Standards based instrumentation and management
– Low overhead and extensibilityCASCON 2007 Presentation by Tosic, Liu, and Zhu The imagination driving Australia’s ICT future
Slide 14 of 24
ASF Architecture Principles
Probes Monitors
Actuator(Middleware Configuration)
Middleware Platform (e.g J2EE Servers,
Enterprise Service Buses, Message Brokers etc.)
ApplicationQoS Requirement
ApplicationStates
Middleware States and Information
Analyzer
Application Logic
Application NICTA Adaptive Middleware
Trade-off computation
CASCON 2007 Presentation by Tosic, Liu, and Zhu The imagination driving Australia’s ICT future
Slide 15 of 24
Existing Middleware (.Net Framework/JAVA EE/ Enterprise Service Bus)Existing Middleware (.Net Framework/JAVA EE/ Enterprise Service Bus)
ASF Architecture Details
Quality Requirement
Policy Definition Policy Definition
ToolsTools
Framework and Framework and
programming programming modelsmodels
Components and Components and
servicesservices
Adaptivity
NICTA Adaptive NICTA Adaptive
Middleware PlatformMiddleware PlatformAdaptive Server
Frameworks
StandardStandard--based Management Layer (JMX/ARM/WSDM)based Management Layer (JMX/ARM/WSDM)StandardStandard--based Management Layer (JMX/ARM/WSDM)based Management Layer (JMX/ARM/WSDM)
Operating SystemOperating System
Engine
ExecutorEffector
Sensor 1
Effector
Sensor n Monitor Analyzer Repository
PolicyManager
Event Correlation
Security Manager
Web Services Wrapper
ASF service layer
CASCON 2007 Presentation by Tosic, Liu, and Zhu The imagination driving Australia’s ICT future
Slide 16 of 24
Use of Interception
• ASF modules (components) are deployed as
interceptorsinterceptorsinterceptorsinterceptors
Interceptor 1
Interceptor 2
Interceptor n
Container
Container
InvocationContext
Target Components
ASF Sensor
ASF Effector
Interceptor 1
Interceptor 2
Interceptor n
Container
Container
InvocationContext
Target Components
Existing (COTS) Middleware Augmented with ASF Components
ASF Policy Manager
ASF Analysers
ASF Monitors
‘plug-and-play’ on
multiple platforms• .Net (ASP.Net and
WCF)
• Axis II SOAP engine
• JBoss App Server
• Mule ESB
CASCON 2007 Presentation by Tosic, Liu, and Zhu The imagination driving Australia’s ICT future
Slide 17 of 24
Case Study: Adaptive Image Server
ClientApplication
Sever
Image DB
1. request a image with
file name,
minimal resolution,
maximum resolution
2. retrieve the image
saved in the BLOB
field
3. scale the image
4. return the scaled
image back to client
•Simulating TPC-W image servers– Based on server workload and network speed, adaptively return images at different levels of resolution and quality to both meet client requirements and optimize performance under peak load
CASCON 2007 Presentation by Tosic, Liu, and Zhu The imagination driving Australia’s ICT future
Slide 18 of 24
Derived Control Loops
Derived control loop for scaling images
Derived control loop for controlling CPU usage
Throughput Response Time
Image Size
Image Scaling Time
Resolution Quality
Network Delay
embedded an
analytical performance model
7/21/2009
4
CASCON 2007 Presentation by Tosic, Liu, and Zhu The imagination driving Australia’s ICT future
Slide 19 of 24
Case Study Implementation
•Case study implemented using ASFimplemented using ASFimplemented using ASFimplemented using ASF and deployed
on JBoss Server 4.0.1 with JRE 1.5.
• Two control loops with eight main customized
adaptive components– ImageScaleEngine
– BandwidthSensor
– ImageScaleMonitor
– ImageScaleAnalyzer
– ImageScaleEffector
– CPUSensor
– ConfigurationMonitor
– ConfigurationEffector
CASCON 2007 Presentation by Tosic, Liu, and Zhu The imagination driving Australia’s ICT future
Slide 20 of 24
Image SizeImage SizeImage SizeImage Size NonNonNonNon----adaptive TPSadaptive TPSadaptive TPSadaptive TPS Adaptive TPSAdaptive TPSAdaptive TPSAdaptive TPS % Improvement% Improvement% Improvement% Improvement
Small (10KB)Small (10KB)Small (10KB)Small (10KB) 54.2554.2554.2554.25 89.2389.2389.2389.23 64.48%64.48%64.48%64.48%
Small Medium (10~100KB)Small Medium (10~100KB)Small Medium (10~100KB)Small Medium (10~100KB) 5.785.785.785.78 8.958.958.958.95 54.84%54.84%54.84%54.84%
Medium (100~500KB)Medium (100~500KB)Medium (100~500KB)Medium (100~500KB) 1.941.941.941.94 3.103.103.103.10 59.79%59.79%59.79%59.79%
Large (500KB~2MB)Large (500KB~2MB)Large (500KB~2MB)Large (500KB~2MB) 0.250.250.250.25 0.310.310.310.31 24%24%24%24%
90% response time is lessthan 23s
90% response time is less than 36s
Adaptatio
n enabled
Adaptatio
n disabled
Evaluation
Simulated workload with peak workload as 65 clients.
CASCON 2007 Presentation by Tosic, Liu, and Zhu The imagination driving Australia’s ICT future
Slide 21 of 24
Some ASF Limitations
•Current specification of policies in ASF is limitedspecification of policies in ASF is limitedspecification of policies in ASF is limitedspecification of policies in ASF is limited
– Design goal was simplicity
– Many details (and special cases) about management goals and actions cannot be expressed
• No supportNo supportNo supportNo support for business-driven management
– Current focus is on maximization of performance
– Conflicting policies selected based on priorities
•But: complements WScomplements WScomplements WScomplements WS----Policy4MASCPolicy4MASCPolicy4MASCPolicy4MASC limitations
– Has broader applicability (not only Web services)
– Platform independent (J2EE, .NET)
– Stronger support for autonomic computing
CASCON 2007 Presentation by Tosic, Liu, and Zhu The imagination driving Australia’s ICT future
Slide 22 of 24
Adding WS-Policy4MASC to ASF
•Goal: leverage compatibilities/synergiesleverage compatibilities/synergiesleverage compatibilities/synergiesleverage compatibilities/synergies to better support businessbusinessbusinessbusiness----driven autonomic managementdriven autonomic managementdriven autonomic managementdriven autonomic management
– For service-oriented (but also other) systems
•Work in progress!
•WS-Policy4MASC is extended with application application application application server configuration actionsserver configuration actionsserver configuration actionsserver configuration actions (easy to do)
• ASF is extended with modules for:
– policy storagepolicy storagepolicy storagepolicy storage (reused from MASC – possible due to loose coupling in both MASC and ASF)
– policy conflict resolutionpolicy conflict resolutionpolicy conflict resolutionpolicy conflict resolution (reused from MASC)
– policy applicationpolicy applicationpolicy applicationpolicy application (new modules are needed)
CASCON 2007 Presentation by Tosic, Liu, and Zhu The imagination driving Australia’s ICT future
Slide 23 of 24
Conclusions and Future Work
• Change is inevitable, adaptabilityChange is inevitable, adaptabilityChange is inevitable, adaptabilityChange is inevitable, adaptability (with minimal human intervention) is needed!is needed!is needed!is needed! => Autonomism
• Execution of IT systems should be alignedalignedalignedaligned with businessbusinessbusinessbusiness objectives, values, and operations
• ASF is a middleware frameworkmiddleware frameworkmiddleware frameworkmiddleware framework to make building application-specific adaptivity simpler
• Adding WS-Policy4MASC to ASF is a step towards a step towards a step towards a step towards businessbusinessbusinessbusiness----driven autonomic computingdriven autonomic computingdriven autonomic computingdriven autonomic computing
•The immediate future workfuture workfuture workfuture work is implementing the new ASF modules to support WS-Policy4MASC policy application (execution)
CASCON 2007 Presentation by Tosic, Liu, and Zhu The imagination driving Australia’s ICT future
Slide 24 of 24
Discussion
•Questions?
• Criticisms?
• Insights?
• Contact information:
•vladat at_server: computer.org (please, start
Subject line with “Your Research” or “IGI Book”)
7/21/2009
5
CASCON 2007 Presentation by Tosic, Liu, and Zhu The imagination driving Australia’s ICT future
Slide 25 of 24
Call for Book Chapters
• “Information Technology Aligned with Business
Objectives and Values: Integrating Software
Engineering, System Management, and Governance”,
edited by Vladimir Tosic
–Value-based software engineering (VBSE)
–Business-driven IT management (BDIM)
– IT governance
– Integration of results from these 3 areas
• Chapter proposal deadline: March 31, 2008
• Chapter submission deadline: April 30, 2008
CASCON 2007 Presentation by Tosic, Liu, and Zhu The imagination driving Australia’s ICT future
Slide 26 of 24
Use of WS-Policy4MASC in MASC
• Feasibility of WS-Policy4MASC demonstrated by
implementing MASC in-memory policy repository,
algorithms for policy decision, and other modules
–Used ‘XML schema to classes’ generator for policy
repository classes
• Examined expressiveness, effectiveness, and
usefulness of WS-Policy4MASC on 2 case studies
(weather report, stock trading)
• To adapt a Web service composition at run-time, only
change WS-Policy4MASC policies
CASCON 2007 Presentation by Tosic, Liu, and Zhu The imagination driving Australia’s ICT future
Slide 27 of 24
Integrating Design- and Run-Time
• Linking all life-cycle stages with bi-directional information transfer, e.g., for impact analysis
• New closed-control loops for adaptation
1. Engineering IT systems during design-time to maximally support run-time management
2. Managing IT systems during run-time to align with and leverage business value information captured in augmented design-time models
3. Feeding back monitored run-time information about business value compliance and various changes to improve/update design-time models
CASCON 2007 Presentation by Tosic, Liu, and Zhu The imagination driving Australia’s ICT future
Slide 28 of 24
UML Profiles for WS-Policy4MASC
• An approach to VBSE+BDIM integration
1. Engineering: Specification of business value (and other non-functional information) in UML models– E.g., using Eclipse plug-ins
2. Management: Generation of WS-Policy4MASC policies from UML models (semi-automatically)– E.g., using XSLT on XMI representations of UML
3. Feedback: Showing run-time management information (e.g., measurements) in UML models (cf. dashboards) and using it for various analyses– E.g., using Eclipse plug-ins
CASCON 2007 Presentation by Tosic, Liu, and Zhu The imagination driving Australia’s ICT future
Slide 29 of 24
UML Profile for Policy Assertions
CASCON 2007 Presentation by Tosic, Liu, and Zhu The imagination driving Australia’s ICT future
Slide 30 of 24
Example Use of This UML Profile
• Part of a case study used to evaluate our UML profiles
• Ongoing work: Tool support for showing run-time management information and corresponding analyses
7/21/2009
6
CASCON 2007 Presentation by Tosic, Liu, and Zhu The imagination driving Australia’s ICT future
Slide 31 of 24
Directory Directory
and Security and Security
ServicesServicesExistingExisting
ApplicationsApplications
and Dataand Data
BusinessBusiness
DataData
DataData
ServerServerWebWeb
ApplicationApplication
ServerServer
Storage AreaStorage Area
NetworkNetwork
BPs andBPs and
ExternalExternal
ServicesServices
WebWebServerServer
DNSDNSServerServer
DataData
Dozens of systems and applications
Hundreds of components
Thousands of tuning
parameters
Middleware
Behind the Scene of Complexity
diagram from Jeff Kaphart’s talk on Software Agents and the Information Economy
Why Adaptive Middleware?
CASCON 2007 Presentation by Tosic, Liu, and Zhu The imagination driving Australia’s ICT future
Slide 32 of 24
Adding Adaptation to Software
• Separation of Concern
• Business logic and are loosely-coupled/independent
• Easy to reconfigure and incorporate new adaptation
• Policy-driven behavior dynamically adapts application to environmental changes
• Adaptive middleware provides programming models and services
CASCON 2007 Presentation by Tosic, Liu, and Zhu The imagination driving Australia’s ICT future
Slide 33 of 24
Fit Into AC Architectural Blueprint
Element
Analyze Plan
Monitor ExecuteKnowledge
Sensors Effectors
Application Server
Analyze Plan
Monitor
ExecutePerformance
Knowledge
Sensors Effectors
Interceptor
Request handler
Proxy
Interceptor
Request handler
Proxy
Standard-based Management Layer
Reflection Reflection
Parameters
Constraints
Correlator
Filter
System model
Control model
Policy management
Concurrency
management
Bootstrap
components
An Autonomic Element
CASCON 2007 Presentation by Tosic, Liu, and Zhu The imagination driving Australia’s ICT future
Slide 34 of 24
Scientific Challenges• The creation of adaptation capabilities presents the following scientific challenges:– the invention of predictive models •to be solved to produce optimal parameter settings for controlling various aspects of application configuration and behavior.
– the invention of new extensible software architectures•to provide flexible application server monitoring and feedback mechanisms, with minimal intrusion.
– the invention of mechanisms to enable dynamic monitoring and adaption policies•reflection mechanisms•policies can be specified declaratively by the application designer.
CASCON 2007 Presentation by Tosic, Liu, and Zhu The imagination driving Australia’s ICT future
Slide 35 of 24
NICTA Extensible Architecture for NICTA Extensible Architecture for Adaptive Application ServersAdaptive Application Servers
ASF Architecture Overview
Application ServerApplication Server
JMX BusJMX BusJMX BusJMX Bus
Operating SystemOperating System
Application Components
Configuration
Business goals to be achieved
by adaptive behaviors
Engine
Policy
ExecutorEffector
Sensor
Effector
Sensor Monitor Analyzer Repository
PolicyManager
CASCON 2007 Presentation by Tosic, Liu, and Zhu The imagination driving Australia’s ICT future
Slide 36 of 24
public void reloadInfo();
public void enable();
public void disable();
public boolean isEnabled();
<<ASF Interface>>
Lifecycle
public String getType();
public String getName();public ElementInfo getInfo();
public void setInfo(ElementInfo info);
<<ASF Interface>>
State management
public void handleMessage(Message msg);
<<ASF Interface>>
Message Consumer
<<ASF Interface>>
Message Producer
public void addMessageConsumer(String messageName,
MessageConsumer consumer);
public void addMessageConsumers(String messageName, ArrayList consumers);
public void removeMessageConsumer(String messageName,
MessageConsumer consumer);
public ArrayList getAllMessageConsumers();
public void removeAllMessageConsumers();
public ArrayList getMessageConsumerByMessageType (String messageName);
public ArrayList getAllMessageTypes();
public void dispatchMessage(Message message, MessageConsumer consumer);
<<ASF Component>>
Sensor
<<ASF Component>>
Monitor
<<Adaptive Component>>
BandwidthSensor
<<Adaptive Component>>
ImageProcessingMointor
public Future submit(Runnable task);
public Future submit(Callable task);
public void execute(Runnable task);
<<ASF Interface>>
Concurrency<<ASF Service>>
ThreadQueuedExecutor
public void bindToExecutor(Executor e);
public void unbindToExecutor(Executor e);
<<ASF Interface>>
ExecutorClient
public void handleNotification(Notification notification);
<<ASF Interface>>
BootStrap<<ASF Component>>
Engine
<<ASF Component>>
ImageProcessingEngine
dispatch task to
init
iali
ze a
nd e
nab
le inst
ance
dispatch task to
init
iali
ze a
nd e
nab
le inst
ance
<<ASF Component>>
ConfigurationEffector
public void addPolicyToScope(Policy p);
public Decision requestGuidance(Policy p);
public Policy resolveConfliction(Policy p,
ArrayList enabledPolicyList);
private String policyName;
private String policyScope;
private String policyDescription;
private boolean policyEnabled;
private ArrayList Conditions;
private Condition policyCondition;
private Decision policyDecision;private int policyPriority;
public ArrayList getEnabledPolicies();
public ArrayList getDisabledPolicies();
public ArrayList getAllPolicies();
public void attachPolicy(Policy p);
public void dettachPolicy(Policy p);
<<ASF Interface>>
PolicyActuator
init
iali
ze a
nd
enable
in
stan
ce
<<ASF Service>>
PolicyManager
initialize and
enable instance
request guidance from
<<ASF Component>>
Policy
<<ASF Component>>
PolicyUtility
<<Adaptive Compoennt>
MessagingTask <<ASF Component>>
Tasks
Customized,Application specific
Adaptive component
ASF pre-defined component with
default implementation
ASF Components and Interfaces
7/21/2009
7
CASCON 2007 Presentation by Tosic, Liu, and Zhu The imagination driving Australia’s ICT future
Slide 37 of 24
Message-based Communication
CASCON 2007 Presentation by Tosic, Liu, and Zhu The imagination driving Australia’s ICT future
Slide 38 of 24
GT4 WS container
ASF Distributed Architecture
ASF Component
Web service
wrapper
Stateful
Resource properties
ASF Component
GT4 WS container
CBE message
can be automated
generated by annotating
ASF components
CASCON 2007 Presentation by Tosic, Liu, and Zhu The imagination driving Australia’s ICT future
Slide 39 of 24Increasing Sophistication
Code
generation Monitoring
Ontology-based
Knowledgerepository
CASCON 2007 Presentation by Tosic, Liu, and Zhu The imagination driving Australia’s ICT future
Slide 40 of 24
Integration with Analysis Intelligence
∧
λ
0λ
maxX
I
W≤=λ
maxX
Self-Management in ASF
CASCON 2007 Presentation by Tosic, Liu, and Zhu The imagination driving Australia’s ICT future
Slide 41 of 24Application to Lending Processes
CASCON 2007 Presentation by Tosic, Liu, and Zhu The imagination driving Australia’s ICT future
Slide 42 of 24
Service Scalability and Availability
7/21/2009
8
CASCON 2007 Presentation by Tosic, Liu, and Zhu The imagination driving Australia’s ICT future
Slide 43 of 24
Observations
1 0.75 0.5 0.25 01 0.9
0.8
0.7 0
.6 0.5 0
.4 0.3 0
.2 0.1
0
20
40
60
80
100
120
140
Siz
e (
kB
)
Quality
Res
olutio
n120-140
100-120
80-100
60-80
40-60
20-40
0-20original size = 57KB
1 0.75 0.5 0.25 01 0.9
0.8
0.7 0
.6 0.5 0
.4 0.3 0
.2 0.1
0
500
1000
1500
2000
2500
Siz
e (
kB
)
Quality
Reso
lutio
n
2000-2500
1500-2000
1000-1500
500-1000
0-500original size = 1.53MB
1 0.75 0.5 0.25 01 0.9
0.8
0.7 0
.6 0.5 0
.4 0.3 0
.2 0.1
0
100
200
300
400
500
600
700
800
900S
ize
(k
B)
Quality
Res
olutio
n800-900
700-800600-700500-600400-500300-400
200-300100-2000-100original size = 614KB
1 0.75 0.5 0.25 01 0.9
0.8
0.7 0
.6 0.5 0
.4 0.3 0
.2 0.1
0
50
100
150
200
250
300
Ima
ge S
calin
g T
ime
(ms)
Quality
Resolu
tion
250-300
200-250
150-200
100-150
50-100
0-50original size = 57KB
1 0.75 0.5 0.25 01 0.9
0.8
0.7 0
.6 0.5 0
.4 0.3 0
.2 0.1
0
100
200
300
400
500
600
700
800
900
1000Im
ag
e S
calin
g T
ime(m
s)
Quality
Res
olutio
n900-1000800-900700-800600-700500-600400-500300-400200-300100-2000-100
original size = 1.53MB
1 0.75 0.5 0.25 01 0.9
0.8
0.7 0
.6 0.5 0
.4 0.3 0
.2 0.1
0
100
200
300
400
500
600
Imag
e S
calin
g T
ime(m
s)
Quality
Res
olutio
n500-600
400-500
300-400
200-300
100-200
0-100original size = 614KB
(a) Scaled Image Size
(b)Image Scaling Time
CASCON 2007 Presentation by Tosic, Liu, and Zhu The imagination driving Australia’s ICT future
Slide 44 of 24
Example Application of ASF
•Enabling Adaptive Web services for
Mobile Applications
– Rendering images based on mobile
application constraints: connection
speed & screen resolutionPortal Container
WSRP Request
Markup Response
SO
AP R
equest
SO
AP R
esponse
ASF ASF
ComponentsComponents
CASCON 2007 Presentation by Tosic, Liu, and Zhu The imagination driving Australia’s ICT future
Slide 45 of 24
Integration with Analysis Intelligence
Layered Architecture of ASF
Operating System (Platform Layer)
Existing Application Server Core Components (Application Server Layer)
JMX Components (Management Layer)
Sensor
Effector
Monitor Analyzer
Engine
Rule
Repository
policy
ApplicationApplication
Policy Manager
Adaptive Component Layer
waiting queue
idle threads
Execution Thread Pool
Task1 Task2
Control flow Data flow